基于標記控制區域生長法的單木樹冠提取1)

甄貞 李響 修思玉 趙穎慧 魏慶彬

(東北林業大學，哈爾濱，150040) (浙江慧鵬地理信息技術有限公司) (東北林業大學) (黑龍江省環境監測中心站)

?

基于標記控制區域生長法的單木樹冠提取1)

甄貞 李響 修思玉 趙穎慧 魏慶彬

(東北林業大學，哈爾濱，150040) (浙江慧鵬地理信息技術有限公司) (東北林業大學) (黑龍江省環境監測中心站)

根據2009年激光雷達數據、正射影像圖及二類調查數據，選取涼水國家自然保護區針葉林和闊葉林樣地進行單木樹冠提取研究，包括利用動態窗口局域最大值法對單木位置進行探測，以及采用標記控制區域生長法進行樹冠邊界的勾繪，分別從樣地和單木兩個層次進行評價。結果表明：樣地尺度上，針葉林和闊葉林的樹冠面積相對誤差的平均值分別為8.74%和-8.24%。單木尺度上，針葉林樣地的生產者精度在62.2%～77.3%浮動，用戶精度在71.5%～83.9%浮動；而闊葉林樣地的生產者精度達到76.1%～91.2%，用戶精度達到78.5%～92.5%。闊葉林樣地勾繪精度浮動較大，但略優于針葉林樣地的勾繪精度，是由于闊葉林樣地中樹冠分布較稀疏所致；而“位置匹配但過度生長”的情況過多是針葉林樣地樹冠提取精度不高的主要原因。

局域最大值；區域生長法；單木樹冠提取；激光雷達；涼水國家自然保護區

Based on LiDAR, orthoimagery and forest inventory data in 2009, we delineated individual tree crown in coniferous and deciduous plots in Liangshui National Nature Reserve, including treetop detection using local maximum method with variable window size, and individual crown boundary delineation using marker-controlled region growing method. The accuracies were assessed on both plot and individual tree levels. For crown boundary delineation, relative error of crown area (RE_CA) of coniferous and deciduous plots were 8.74% and -8.24%, respectively. Producer’s accuracy (PA) and user’s accuracy (UA) for conifers were 62.2%-77.3% and 71.5%-83.9%, while PA and UA for deciduous plots were 76.1%-91.2% and 78.5%-92.5%, respectively. The deciduous forests had more fluctuant but higher PA and UA than conifers due to the lower canopy density. The case of “match but over growing” was the main reason that the accuracy was low in coniferous forests.

樹冠是樹木獲取光能、進行光合作用的重要場所，樹冠在預測樹木生物量、監測樹木長勢、判定木材材性等方面也起著重要作用。同時，對通過樹冠傳遞的各種能量質量模型來說，樹冠是個重要的構建因子。……